How To Analyze Customer Feedback
Your customers go out of their way to explain their difficulties – but are you properly equipped to hear them?
What is customer feedback?
First, let’s define what customer feedback actually is. In its simplest form, customer feedback is the qualitative or quantitative opinions + user experiences your customers share with you.
Feedback can differ greatly, so it’s helpful to view it through a series of scales. Customer feedback can be:
- Positive or negative... or neutral
- Quantitative or qualitative
- Critical or constructive (ie. complaints vs. requests)
- Specific problems or general emotional sentiments
These are important distinctions because they all serve different purposes (more on this in a future blog!). Being able to capture feedback from each category is key to ensuring that customer issues 1. aren’t being missed and 2. are routed to the correct teams for action.
Tangibly, customer feedback is often captured through an omnichannel customer feedback collection program. These programs usually include channels like:
- Phone support via call centers
- Live chat with agents
- AI chatbots (pre-programmed with auto-solving knowledge bases)
- NPS surveys (Net Promoter Score)
- Social media (support and social listening)
- Reviews (eg. Trustpilot, App Store / Play Store, or product reviews)
- SMS
- CSAT & DSAT
- Online forums (Reddit etc.)
What are the popular methods to analyze customer feedback?
Since there are so many ways to slice and dice data these days, it’s important to have a goal in mind of what you’re looking for before you go digging. Moreover, it’ll help to define a purpose for your analysis – are you looking for product behavior insights? Trying to forecast user adoption in a specific region? Or perhaps you’re hunting down software bugs and customer experience difficulties? These are all very different goals and there are different tools and strategies you can employ for each.
Data Visualization: tools like Power BI or Tableau provide a great holistic overview of many significant metrics of your business. For example, Tableau has recently expanded their CX analytics dashboard capabilities to include quantitative speech analytics gleaned from call center operations.
Sentiment Analysis: tools like BrandWatch or Talkwalker’s “Quick Search” focus on your customers’ feelings toward your brand on social media. Tools like Chattermill are in the mix as well, helping to disposition feedback from customer conversations.
Text Analytics: This is a broad term that refers to many applications of NLP (Natural Language Processing), but we’ll focus on some of the main applications for NLP in customer feedback data, specifically. The goal of course is to glean quantitative insights, trends, and patterns within freeform text. AI helps us to do this automatically and at scale, instead of manually sifting through data. Many software companies today offer some level AI-based text analytics at varying levels of complexity and for different use cases when conducting customer experience analysis, and I’ll feature a few below.
How are text analytics models used to analyze customer feedback?
There’s 4 popular models that are utilized to analyze text today – word clouds, topic modeling, classifiers, and key phrases. They each identify certain patterns, but are limited in what they can find.
Word clouds
What it does: displays keywords and their frequency of occurrence
Limitations: word clouds aren’t insights on their own
Topic modeling
What it does: word frequency combined with probability math to discover meaningful topics
Limitations: insights aren’t specific enough to be actionable
Classifiers
What it does: groups data into meaningful buckets
Limitations: can’t discover new insights & emerging issues; can't find new classes
Key phrases
What it does: finds phrases using grammar rules and groups them together
Limitations: Doesn’t work for customer support data; wouldn't pick up similar context
These models have significant failure modes
These popular text analytics models were originally intended to carry out simple tasks, like assigning tags to news articles to categorize them, but they’re not going to be able to pull descriptive insights from support conversations. As far as NLP applications go, their ranges are very limited.
As mentioned in another blog, most companies that offer text analytics (even the bigger ones like Medallia, Qualtrics XM, Oracle, or Clarabridge) don’t attempt to process feedback channels other than NPS, surveys, or social media, because unstructured data in customer support conversations is just too complex a challenge for their text analytics models to tackle.
Their respective failure points include, (but are not limited to):
- Vulnerability to bad grammar
- Ignorance of semantics
- Inability to find new topics, (or “unknown unknowns,”)
- Uninterpretable results (eg. lists of random keywords with no context).
If you want to find issues buried in customer feedback, these are baseline requirements you must solve for.
More to the point, these models are rigid and are incapable of handling “niche” products or your company’s unique vernacular and new product names because they only look for keywords and phrases that appear in pre-defined buckets. Lots of text analytics companies utilize models that are tuned to, say, ‘popular words in banking.’ These models are typically very generic, aren’t tunable to your customers’ specific voices, and won’t pick up on feedback when it is given imperfectly, which…is most of the time! And how can you be confident in the accuracy of results if these popular models automatically exclude all imperfect data? (You can’t 😅)
So, what model can handle the demands of messy customer feedback?
Customer feedback is constantly evolving, as is your product. New hardware & software come out daily, you’re shipping new features + integrations on a regular basis – there are endless cracks through which your CX can slip, so it’s increasingly important to have a machine learning solution that can automatically detect and map it all out for you in real time.
In order to accurately analyze customer feedback today, you need an adaptive, elastic model that constantly improves and iterates at the speed of your customers. I’ll say it again, louder for the people in the back! 😃👏
"In order to accurately analyze customer feedback today, you need an adaptive, elastic model that constantly improves and iterates at the speed of your customers"
Gone are the days when you could manually read all of your customer feedback.
Gone are the days when rigid text analytics models were our only option.
Gone are the days when your customers were willing to suffer through a poor customer experience – today customers will just leave. Even after a detailed churn analysis, they’ll be nearly impossible to recapture. This is why it’s important to keep your customers happy while you have them.
What is Spiral?
Specific customer issue detection using the latest machine learning (ML) + AI research, and neural networks.
What’s the Spiral experience like?
Because we thoroughly analyze customer support channels, each time you log into Spiral, we present you with the full list of current customer issues, starting with the biggest issue at the top, so you can decide what to work on and in what order.
You can also subscribe to alerts for emerging issues or to receive ongoing reports of any number of issues or categories.
If your customers talk about it, Spiral ensures you know about it.
How can I try Spiral?
If you’re curious what Spiral’s turnkey insight solution can find in your customer feedback, let’s chat. Schedule a demo and before our first call, we’ll scan your company’s customer reviews and show you a personalized demo dashboard based on your public data.
We can get started with as little as one feedback channel, or process all channels in your omnichannel support program. We work with live chat, support emails, phone calls, NPS, DSAT, CSAT, social media, reviews, SMS, surveys – or any channel you consider to be feedback.
Our customers come from spaces such as: financial services / banking / credit unions, fintech, connected hardware and software, insurance, retail, B2C SaaS, and managed marketplaces – and we’d love to hear from you.
Thanks for reading!
